First let me show you some pictures of these particular headphones, and later, we'll get into measurements with the inexpensive miniDSP (H)EARS test fixture.In the image of the opened AKG K371 above, we see that it comes as a rather complete package. You can detach the headphone cable from the drivers easily with its small locking DIN-type connector that comes out of the left earcup (the cable connector has a spring push-button to unlock):
I might just have to stick a piece of red/blue tape on the outside (color coded like with Sony studio headphones) to remind me of which side is which! Overall, I find the headphone quite sturdy and should not break easily assuming it's not abused.
I. Headphone Measurements using the miniDSP (H)EARS...
Let's talk about measuring headphones for a bit and introduce you to the miniDSP (H)EARS. Most of the time the "H" is dropped from this measurement jig although on the front of the unit, we see the label Headphone & Earphone Audio Response System which I think is a nice acronym:
|Front DIP switches used to change microphone gain. I'll just keep it at the default +18dB.|
If you have one of these, be mindful of the 4 screws holding down the synthetic ear/head in place; these areas of depression can make it more difficult to create a good seal. I know some folks have removed the screws and I've also had some success with using thin foam tape (like the soft skin-hugging tape used in Band-Aids) to smooth out the irregularities as well. Generally, with a bit of care, I have not found this to be a big problem. The higher density and lower compliance of the pinnae will also make it a bit more difficult to measure on-ear (supraaural) headphones with a proper seal however, leading to inaccurate poor bass response. Getting good, consistent on-ear headphone measurements has been by far the most difficult task I've tried thus far (hint: use some large, thick rubber bands around supraaural headphones to clamp down on the pinnae helps).
Another area of concern is that the ear canal is just a straight tube leading to the measurement microphone inside. The canal is also rather short. Human external auditory canals are more complicated than this which would affect the resonance characteristics leading to the tympanic membrane. Be careful with IEMs, especially those with longer tips that could jam against the microphone. Regardless, a compensation curve will need to be applied to correct for the resonances (miniDSP includes their IEM Diffuse-Field "IDF" compensation which we'll look at another time).
I can certainly understand Brent Butterworth's concerns and article about the miniDSP EARS back in 2018. Nonetheless, I do believe the inexpensive US$199 price has indeed brought "headphone measurement capability to the masses" and opened up the opportunity for the hobbyist audiophile to explore relative sound quality of headphones without aiming for absolute "industry standard" results.
It's interesting when we look around at headphone measurements online and notice just how variable the frequency response graphs presented are among sites like In-Ear Fidelity/Crinacle, InnerFidelity, DIY-Audio-Heaven, Rtings, Reference Audio Analyzer, and Audio Science Review. Unlike speakers where we can generally agree on measurements based on how a speaker would perform anechoically, headphone performance by definition is affected by the interface or "coupling" between it and whatever measurement system it's being applied to, whether just a flat plate, some kind of simulated "ear" system, or even live human subjects in some instances (I believe Rtings does this, check out their "Challenges & Solutions" video). The variability among humans and measurement techniques are also well laid out in this 2008 Stereophile article by Keith Howard.
By the way, do not forget the variability between the same model of headphones as well! Quality controls will affect the variability between each headphone made. Also, it seems that some manufacturers can change headphone characteristics even within the same product line - for example, check out the variability with the Campfire Andromeda IEM.
It's tempting to look at the prices for some of these Head And Torso Simulators (HATS) and ear-simulators thinking that one needs this level of equipment in order to test headphones. However, with all these variables, the "circle of confusion" when it comes to headphone measurements is relatively large. As such, I think it's important to not obsess on an exact graph or single result and recognize that there is only so much actual precision we need be concerned about and there is an element of "skill" that needs to be developed by whomever is running the measurements in achieving the consistency using whatever equipment one has (something I'm in the process of building through gradual experience with this stuff). There is a point where the measurement accuracy is also "good enough" as hobbyists and consumers. Look for trends and understanding of the broad strokes of what the measurements are telling us...
Let's talk about the most important measurement - frequency response. Specifically, how should we display the frequency response graph?
The problem is that since the headphone has to couple with human ears, we all have our own head related transfer function (HRTF or Anatomical Transfer Function) which filters the sound that hits the tympanic membrane (microphone in the case of the measurement device) and this is in principle where we are measuring for an "accurate" sound. In fact, have a look at your own ears in the mirror. I bet you'll notice some asymmetry in the size and folds of the pinna - I can certainly see and feel these differences myself. It's remarkable how the brain through experience ultimately filters out the sound to make the stereo tonality perceived as "balanced"!
Since no two ears are exactly alike, what we can say is that the anatomical differences will lead to variation with usually a resonance peak between 2-5kHz. For a typical ear, around 3kHz one might see up to +17dB! In the same way, the measurement jig with its synthetic ears will create a unique resonance pattern which might or might not conform to the resonance pattern of whatever "average" ear we ascribe to humans.
Logically, and simply being consistent with other device measurements, "ideal" should be aiming for a flat target graph IMO. Reading a "raw" graph is fine, but I don't see any benefit to readers "knowing" the raw shape since ultimately it's asking the reader to estimate variance from some "average" anyway. Therefore, when a frequency response is measured, how do we take this into account and convert to a "flat" looking standard curve that correlates with human preference for good sound? With what "compensation" curve then would we use?
Over the years, there have been free-field and diffuse-field targets of multiple forms applied to headphone measurements (see this Audio Precision note to see typical shapes of these). More recently, the Harman targets, particularly the Olive-Welti-Khonsaripour (2018) paper "A Statistical Model that Predicts Listeners' Preference Ratings of Around-Ear and On-Ear Headphones" has been commonly referenced. For us hobbyists, I think Tyll Hertsens' article a few years back before his retirement captures the essence of this complexity down to the level of human ear anatomy and those resonances.
So, the question for me to decide then was: "With the miniDSP EARS, what are you going to do when it comes to headphone measurements on this blog!?"
Well, until a better target comes around, let me stick with referencing the empirically-derived Harman 2018 AE/OE Target (as per the link above). For more on the Harman curve, check out Headphonesty's 3-part article, impressive work.
Keeping this next part short and sweet even though it took me about 3 weeks to get the job done to a reasonable level of satisfaction, broadly, here are the steps I took to create a compensation curve for my (H)EARS that I could be happy enough with:
1. Start with the miniDSP-supplied compensation curves called "HPN" and "HEQ", measuring a bunch of my headphones to see how these compare to other results posted online. It's quite clear looking at the "raw" results that the EARS has a strong concha/pinna/canal combined resonance up at 4.2kHz which is higher than the normal human ear resonance. While I like the HEQ's "flat" target in higher frequencies, it doesn't deal with the strong resonance as well as HPN. I ended up using a hybrid of the HPN below 5kHz blending with HEQ above 5kHz. I focused on the left compensation curve as a foundation and adjusted parameters to target some of the components of the Harman 2018 curve like the small sub-bass accentuation.
2. Using the left channel compensation from (1), now measure the same headphone earcup on the RIGHT side to check on right-left relative balance. Since this is the same driver but different silicone ears and microphone, tweak the mirror compensation curve to appear similar to the left. For this, I used my headphones with symmetrical ear cups like the Creative Aurvana SE and AudioTechnica ATH-AD700 - both relatively easy to measure with large enough earcups for proper seal.
3. With the general shape and left-right levels tweaked, do some final adjustments with a larger range of headphones. I drew inspiration from this Super Best Audio Friends thread for example. The AKG Q701 and Sennheiser HD800 are good headphones to use here. They have asymmetrical earcups/pads so not good for step 2, but easy to measure with consistent seal and large circumaural fit for final adjustments.
For your consideration, here is the final "Arch-Hybrid Compensation" (version 1.5) curves, both left and right plotted against the original miniDSP HEQ and HPN, along with Harman-2018 AE/OE:
II. AKG K371 Measurements - Core
Without further ado, let's have a look at the AKG K371 "Core" results:
To the top right, I've plotted the electrical impedance and phase for this headphone (usually I'll just use the left driver unless I see some significant discrepancies between channels). Minimum impedance is around 36.5Ω with a driver resonance peak at 24.5Hz at 42Ω. I've included a sensitivity reading as well with a 1kHz tone at 0.25Vrms. In the case of the K371, it's an extremely sensitive headphone! 106dB SPL/0.25V is approximately 118dB/V assuming good linearity (1Vrms is +12dB over 0.25Vrms). This is close to Hi-Fi News' measurement of 120.6dB/V done a few years ago. The official spec for the K371 is 114dB/V.
Bottom left we see the colorful waterfall plot. I've chosen a setting of 40dB Y-axis range from the peak, 100Hz to 20kHz along X-axis, and 10ms Z-axis timeframe. The waterfall has a nice and steep initial decline with no resonances to be found within 2ms from about 1.5kHz and above. Over time, there's some residual low-frequency resonance that bounces back like an echo that lingers below 1kHz out to 10ms. This is not uncommon with closed or semi-closed headphones I've tested. I wonder if more internal damping might help reduce cup resonance.
Bottom right is a look at the harmonic distortion profile with the headphone playing at my standardized level of 95dB SPL at 1kHz which is louder than my usual listening level. Distortion is plotted relative to fundamental at 100%. I've put the cursor at 1%. Given how forgiving the ears are when listening to music, I'm honestly not too worried about distortion causing much of a problem unless it significantly breeches that 1% threshold through much of the audible spectrum (remember the results of our harmonic distortion blind test a year ago). The K371 is nice and clean with mainly 2nd order harmonics (the amount of 2nd and 3rd harmonics particularly will vary for headphone with measurement jig interaction). We can see a bit of distortion poking above 1.5kHz but it's not much. Notice that I faded out the 3.5-5.5kHz portion. This area is where the EARS measurement jig has that strong 4.2kHz resonance and this artificially worsens the distortion result within that area; again, limitation of the test gear. However, notice that there's what looks like a second hump just above 5kHz which is likely a true area of increased distortion with these headphones.
While I might not do this regularly, we can look at the headphone linearity:
III. AKG K371 Measurements - Isolation and Leakage
Beyond the "Core" measurements, we can look at how the headphones interact with the outside world.
Fist, let's consider noise isolation which is how well, given a good seal, wearing the headphone is able to attenuate outside noise. We can do this comparison of relative noise level when a loudspeaker is playing at a preset SPL with the headphones on or off.
I used my Energy C100 speakers playing white noise at 1' away to create this graph:
As you can see, the headphones provide some isolation from outside noise above 400Hz; achieving about -9dB at 1kHz and around -20 to -25dB by 5kHz using my test methodology. The EARS will pick up noise from behind the headphone so in actual use, the noise isolation will be a bit better. I don't think it makes much sense to show too much detail here so I used 1/3-octave smoothing.
Conversely, we can measure how much headphone leakage there is (ie. how much of the sound can people nearby hear?). Ideally, we should be doing this in a quiet little anechoic chamber/box, but since I didn't feel like building one, here's what I did. Set the headphone output to an average of 100dB SPL as measured by the EARS. Then run a sweep measuring the sound at 1" away from the front-center "seal" between the earpad and measurement rig using an external miniDSP UMIK-1 measurement microphone. For a closed headphone like this, this area of the seal between the face and earpad seems to be where most of the sound is escaping from. It's not hard to put one's ear near the headphones to listen and select an appropriate place to capture a reading. Because of signal-to-noise limits of the UMIK-1, I chose to get closer than a more typical 6" or 1' position. With open headphones, the area of maximal leakage could be just behind the driver in which case I'll grab a reading from there instead. Not difficult to get a couple readings and find an average as well. I do this in a quiet room and also put a thick blanket over the headphone/microphone to reduce outside noise from contaminating the result.
Here's what the leakage spectrum looks like with these AKG K371 headphones:
In terms of look and feel in use, I found the K371 quite comfortable in the hand and on the ears. It's pretty well all black which is good for its intended studio purpose - don't buy these for impressive looking "bling" :-). The rubberized texture is comfortable. They're quite light weight (260g measured without cables) which is great for longer term listening. The plastic is not as creaky as other studio-type headphones I have (Fostex T50RP/Dekoni Blue, AudioTechnica ATH-M50, Sony MDR-V6 come to mind) and I like the soft synthetic ear pads as well which seal better than most headphones I've used but they do get a bit warm compared to the relative comfort of thick velour pads for example.
As for sound quality, I would say that these are quite "pleasant". I listened to them for a couple of weeks before these measurements, and indeed, I thought there was more sub-bass with these than most headphones. Given the closed nature and relatively easy-to-seal soft earpads (so long as you don't have huge ears!), one can appreciate the pressurized low rumble in bass-heavy music like Massive Attack's "Angel" (Mezzanine), Hans Zimmer's "Time" (Inception soundtrack), or the sub-bass-heavy "2049" and almost frightening intensity of "Flight to LAPD" from Blade Runner 2049. For those of us who love the bass experience in a home theater, headphone bass alas does not provide the powerful overall somatic experience of a subwoofer in the room and the thumping against the chest. But it can shake the noggin'. :-)
After running the measurements and noticing the upper-mid dip, I had a good listen to the headphones with a sine sweep and yes, I can hear that ~4kHz dip. I don't think it sounds as pronounced with my ears compared to what is found on testing. I wonder if human ear sensitivity as per the Fletcher-Munson curve has a significant part to play here where we are most sensitive between 3-4kHz which happens to also be the locale of this dip.
With a dip like this, there could be a positive effect of reducing harshness in some music. For example, the K371 does a good job with compressed modern pop. As a catchy tune, I think Dua Lipa's "Love Again" on Future Nostalgia sounds great and is my favourite track on the album. These headphones don't sound "sizzly" which is good for longer term listening on stuff like this. Likewise, that 3-4kHz dip does reduce sibilance and some of the brightness on tracks like the old audiophile acoustic nugget "Keith Don't Go" (Acoustic Live). Apart from this, the treble has adequate "brilliance" when called for.
More typical audiophile fare like female vocals such as Julia Fordham's Collection sounded natural with excellent separation of vocals from the background instruments.
Soundstage isn't expansive like open headphones, the sense of depth and any "outside the head" experience isn't there. While not really a fair comparison given headphone type and price differential, the Sennheiser HD800(S) clearly sounds much more expansive compared to these.
As a pair of studio monitoring headphones, these are great for resolution; listening to those little nuances. It's easy to pick apart poor recordings. Distortions from low-bit lossy codecs sound very obvious and ugly with these. A couple months back, I mentioned the Korean drama It's Okay To Not Be Okay soundtrack and the track "Her World (Moon Young's Theme)". I didn't notice just how much swirling intermodulation-type distortion was in the first 10 seconds (quite possibly intentional and fits with the emotional tone of the track) until I had a good listen with these.
|AKG K371 on miniDSP (H)EARS during measurements. Testing best performed with a shot of XO cognac... :-)|
I hope this post gives you an idea of the kinds of headphone measurements and presentation format I'll likely be using in the days ahead. Also, for those considering using the miniDSP (H)EARS, I hope this touches on some important considerations (here's an excellent review with more info and comparisons). While one could just use the miniDSP-suppled calibration/compensation curves ("HEQ" and "HPN"), I think it's worth taking some time to do some "triangulation" using popular high quality headphones you may have to tweak the compensation curve. Regardless, whatever one does, there will be idiosyncrasies due to the non-standard gear. Comparing measurements using my Sennheiser HD800, AKG Q701, Creative Aurvana SE/Live!, AKG K371, and looking at what others have found is useful when doing some detailed tweaking.
As for the AKG K371 headphones, while I think it's too early to call these "classic" headphones (released in November 2019 so time will tell), I think they make an interesting modern comparison standard when talking about relative performance. They're easily available in the retail channels, relatively inexpensive, and the low ~35Ω impedance along with very high sensitivity measured around 118dB SPL/V at 1kHz makes this headphone very easy to drive even with low-powered mobile gear.
AKG, being a Harman subsidiary, has implemented the "Harman target curve" into these so you'll get a chance to listen for yourself if you like this kind of tonality. As discussed above, they do accentuate the sub-bass a bit. I don't know what the threshold is for those who consider themselves "bass-heads", but these headphones are certainly not bass-light and for me is tastefully done. If you're doing production work, be careful about the 3.7kHz dip and double check with other headphones and speakers to make sure that region doesn't sound too harsh. It think these would make for excellent "workhorse" closed headphone if you're into home recording with reasonable noise isolation and leakage. To put this into some context, when I have these headphones on and listening to music at normal levels, I can't hear my home phone ringing nearby (I have no problem hearing the phone ring while using a semi-open Dekoni Blue/Fostex T50RP, and definitely no issue with hearing what's happening around me with the open AKG Q701 and Sennheiser HD800). As for leakage, when my wife is sitting across from me about 1.5m away, she barely notices me listening to music nor can she hear the silly audiophile YouTube videos I'm listening to. ;-)
If I were to EQ, just a simple parametric 3.75kHz, +6dB, Q5 peak would be adequate.
Alright audiophiles. That's all for this week... Stay safe and hope you're enjoying some sweet tunes!